In special machinery manufacturing a quote costs more than it is worth as long as nobody knows what the last comparable machine actually cost. The bottleneck is not writing the document but the pre-costing, and that almost always fails at the same point: the post-calculation exists, but nobody can find it again. Automation therefore works precisely there, not in template management.
This article walks through the quotation process station by station and separates what can be automated from what stays a decision. The second list matters more.
What does a quote in special machinery manufacturing consist of?
Unlike catalogue goods or spare parts, there is no price to look up. Every enquiry is a small development project that has to be costed before anyone knows whether it is worth having. In practice an enquiry passes through six stations:
- Intake and triage. The enquiry arrives as an email with a specification, often as a PDF, sometimes as a drawing set or a phone note.
- Technical clarification. What does the customer actually want, how much of it is feasible, and which questions must be answered before costing?
- Concept and preliminary design. Enough engineering to be able to cost the job, and not one hour more.
- Pre-costing. Bought-in parts, manufacturing hours, engineering hours, assembly, commissioning, risk margin.
- Quote document. Technical description, price, delivery time, payment terms, warranty.
- Follow-up and negotiation.
Station 5 is the only one that looks like quotation from the outside, and it is the shortest. Anyone wanting to shorten the lead time has to work on 2 through 4.
Where does the time actually go?
| Station | Share of effort | Automatable |
|---|---|---|
| Intake and triage | low | largely |
| Technical clarification | high | partly |
| Concept and preliminary design | high | barely |
| Pre-costing | high | in parts, and this is the lever |
| Quote document | low | largely |
| Follow-up | medium | largely |
The distribution explains why quote configurators so often disappoint in this sector. They attack station 5, the item that costs the least anyway.
What is the real bottleneck?
The question every pre-costing has to answer at its core is this: what did a similar machine actually cost us?
In most companies that information exists. It sits in past quotes, in post-calculations, in project closing reports and in the heads of three to five people. It is simply not findable. Anyone looking for it searches a filing structure that grew over two decades and concludes after twenty minutes that estimating is faster.
That is exactly where the error originates that gets expensive later. Not because someone costs badly, but because they cost without the best available basis.
The test for whether you have this problem
Ask the costing team for the three most similar projects of the past five years and their post-calculations. If the answer takes longer than half an hour or comes from memory, your leverage is not in the quote document.
What can be automated?
Four things pay off regularly in practice.
Capturing and classifying enquiries. Incoming enquiries get read out, assigned to a machine family and tagged with the key data needed for triage. This saves little time per enquiry but ensures a searchable history comes into existence at all.
Finding comparable past projects. The most important point. A system that searches specifications, past quotes and post-calculations by content answers the question from the previous section in minutes rather than hours. It does not produce a costing, it produces referenced comparison cases.
Proposing a pre-costing from past data. From those comparison cases and the post-calculation history, a first estimate can be derived, including the range in which comparable projects actually landed. That range is often more valuable than the average.
Producing the document and following up. Text blocks, technical description from the concept, deadline tracking. Small in effect, but cheap to have.
What stays with people?
Three decisions cannot sensibly be automated, and attempting it does harm.
The feasibility decision. Whether a required cycle time is achievable with the intended principle is not decided on the basis of similarity to past projects. This is precisely where a special machinery builder carries risk.
The risk margin. It depends on the customer, on the maturity of the specification and on how much new ground the machine covers. A model that derives it from history perpetuates past misjudgements.
Pricing strategy. Whether a job is quoted at contribution margin or at market price is a sales decision. It does not belong in a costing system.
When is this not worth it?
At a low win rate, automation amplifies the wrong behaviour. Winning one quote in eight and doubling the process mainly doubles unpaid effort. The more honest first question is then not how quotes get produced faster, but which enquiries should not get a quote at all.
Also not worth it: companies where every machine really is a one-off without precedent. The whole approach stands or falls with the existence of comparable past projects. Fewer than roughly thirty usable cases is not a basis, it is a collection of anecdotes.
And finally: if the post-calculation is never produced in the first place, the decisive data source is missing. That is not a software problem but one of process discipline, and it has to be solved beforehand.
What data does this require?
- Past quotes with technical content, not just the price page.
- Post-calculations for as many of those quotes as possible.
- Order outcome: won or lost, and where possible why. Without this point a system learns how you calculate, but not whether you were right.
- Bills of materials and ERP data as a supplement for bought-in parts and manufacturing times.
The third point is missing most often and is the most valuable. It is also the only one that cannot be obtained retrospectively.
How does this relate to document search?
Both topics draw on the same body of documents but pursue different goals, and that separation matters in practice.
Anyone wanting to search documents, standards and project knowledge and get the answer referenced to file and page is looking for a tool. That is KoAssist, which answers from your own documents and supplies the source location.
Anyone wanting to derive a costing proposal from that and wire the result into ERP and the quotation process needs a development service. That is the subject of this article and what we build.
Want to know whether your past quotes and post-calculations are usable as a costing basis? We look at what you have and tell you what can be derived from it and what cannot.
Book a free intro callHow do you start?
Not with the quote document. The first step is taking stock: how many usable quotes and post-calculations from the past five years exist, how consistent are they, and is the order outcome documented?
That question can be answered in one to two weeks and decides everything that follows. If the answer is negative, the right measure is not an AI project but disciplined post-calculation starting now. In two years the basis that is missing today will be there.
How such a project runs and what it costs is covered in our overview of AI project costs. The related case in after-sales, with a different scope and different data sources, is in Automating spare parts quotes.
FAQ
Can quotation in special machinery manufacturing be automated?
Partly, and at a different point than most people expect. Writing the quote document is the smallest item. The effort sits in the pre-costing, and that is where the leverage is: finding comparable past projects, reading their post-calculations and deriving a defensible estimate. The feasibility decision and the risk margin stay with people.
How long does a quote in special machinery manufacturing usually take?
Between half a day and several weeks, depending on complexity. The bulk goes into technical clarification and pre-costing, not into the document itself. Anyone wanting to shorten the lead time has to start at the costing basis, not at template management.
What is the most common cause of mis-costed quotes?
The post-calculation gets written but never read again. That removes exactly the information that would have prevented the error on the next similar project. The second common cause is engineering hours, estimated as a lump sum in pre-costing and regularly overrun in reality.
Does automation pay off at a low quote win rate?
Only if it does not make the win rate worse. Writing twice as many quotes at a rate of one in eight mainly doubles unpaid effort. The more valuable automation therefore first answers which enquiries are worth pursuing at all, and only then how the quote gets produced faster.
What data does such a system need?
Past quotes, their post-calculations, and the information on which quote was won and which was not. Without the last point the system learns how you calculate, but not whether you were right. Bills of materials and ERP data come in as a supplement.
How does this relate to standards and document search?
Both draw on the same body of documents but pursue different goals. Searching documents and getting an answer referenced to file and page is a product topic and sits with KoAssist. A system that derives a pre-costing from past quotes and is wired into your processes is a development service and the subject of this article.
